A Multilingual, Large-Scale Study of the Interplay between LLM Safeguards, Personalisation, and Disinformation